1. Technical Field
The present invention relates to a method and apparatus for measuring high-frequency electrical characteristics of a two-terminal electronic device, such as a chip inductor, a chip capacitor, a chip resistor, or the like. More specifically, the present invention relates to a method for correcting a measurement error in measuring the impedance or the Q-value of the electronic device with a reflection method using a measuring device, such as a network analyzer or the like.
2. Background Art
To measure high-frequency electrical characteristics of an impedance device, such as a surface-mounted chip inductor or a surface-mounted chip capacitor, using a network analyzer, there is a measurement method in which a planar transmission line (such as a microstrip line or a coplanar waveguide) is connected to the network analyzer via coaxial cables or the like. It is impossible to directly connect the coaxial cables to the electronic device. Therefore, the electronic device is brought into contact with the planar transmission line to make a measurement. In this case, in order to obtain true values of a scattering parameter matrix of the impedance device serving as a test object, it is necessary to identify error factors of a measurement system and to remove the effects of the error factors from the measurement results. This is referred to as correction or calibration.
In the measurement using a network analyzer, as shown in non-patent document 1, Application Note 1287-9; In-Fixture Measurements Using Vector Network Analyzers (©1999 Hewlett-Packard Company), TRL (Through-Reflection-Load) calibration and SOLT (Short-Open-Load-Through) calibration are known as techniques for removing errors of the measurement system.
FIGS. 1 and 2 show respectively SOLT and TRL measurement systems using a network analyzer and the corresponding error models for use in SOLT calibration and TRL calibration.
An electronic device 1 serving as a test object is connected to a transmission line provided on the top surface of a measuring fixture 2. Two ends of the transmission line on the measuring fixture 2 are connected to measurement ports of the network analyzer, which is not shown, via coaxial cables 3.
In the error model of SOLT calibration, S11A, S21A, S12A and S22A are scattering parameters of the transmission line including the test object, EDF, ERF, and ESF are scattering parameters on one measurement port side, and ELF and ETF are scattering parameters on the other measurement port side.
In the error model of TRL calibration, S11A, S21A, S12A and S22A are scattering parameters of the test object, e00, e10, e01 and e11 are scattering parameters on one measurement port side, and f00, f10, f01 and f11 are scattering parameters on the other measurement port side.
In order to identify error factors, it is necessary to fix at least three types of devices (standards) whose scattering parameters are known to a test object measuring plane and make measurements. Traditionally, opens, shorts, and terminations (loads) (=50Ω) are often used. Since these standards can be implemented in a coaxial environment, this method, which is referred to as SOLT calibration, is widely used. In SOLT calibration, as shown in FIG. 3, three types of connectors 4 including a short (0Ω), an open (∞Ω), and a termination (50Ω) are used, and the ports are directly connected to each other to achieve a through state.
However, in the case of SOLT calibration, it is very difficult to implement these standards in environments other than the coaxial environment, and the standards necessary for calibration cannot be fabricated in the form of a chip device. For example, a planar transmission line for use in measuring a surface-mounted device is, unlike a waveguide or a coaxial transmission line, unable to achieve a satisfactory “open” or “termination”, and it is thereby practically impossible to perform SOLT calibration. Also, in general, measured values obtained by measurements are not characteristics of the test object 1 alone, but are composite characteristics of the test object 1 and the measuring fixture 2 to which the test object is connected. It is thus impossible to measure characteristics of the test object alone.
TRL calibration employs, as shown in FIG. 4, instead of device-shaped standards that are difficult to realize, a (through) transmission line 5a whose ports are directly connected to each other, a total reflection (reflection=normally shorted) transmission line 5b, and several other types of transmission lines 5c and 5d of different lengths as standards. With regard to the transmission lines 5a to 5d, it is relatively easy to fabricate transmission lines whose scattering parameters are known. Also, if the total reflection is achieved by shorting, it is relatively easy to estimate characteristics thereof. Therefore, these transmission lines are sufficient to perform calibration. Basically, it is possible to measure the characteristics of the test object 1 alone.
In this example, the through transmission line 5a is a so-called zero-through. To measure the test object, the test object is connected in series with the measuring fixture 2 whose length is greater than the through transmission line 5a by the length of the test object, and a measurement is made.
However, when TRL calibration is applied to a surface-mounted device serving as a test object, the following problems occur.
1) With regard to the transmission lines (several types of lines, including reflection and through lines) 5a to 5d serving as the standards, it is necessary that all the error factors generated in connections between coaxial connectors 3 and the transmission lines 5a to 5d be equivalent. However, even when the same type of connectors are used as the standards, characteristics of the standards vary greatly when the standards are connected to a measuring device, thereby generating calibration errors. It is practically impossible to perform TRL calibration near a millimeter-wave band.
2) In order to solve this problem, the coaxial connectors 3 are common among the transmission lines 5a to 5d, and coaxial pins are in contact and connected to the transmission lines serving as the standards, thereby avoiding the effects of variations in connector measurements. Structurally, however, it is difficult to ensure a sufficient pressing load at the connections, and hence the coaxial pins may be damaged. Since the connections are unstable, calibration becomes also often unstable. The higher the measurement frequency, generally the thinner the transmission lines and the coaxial pins. Depending on the positioning repeatability thereof, measurement variations may become larger.
3) Since it is difficult to determine, in the calibration operation, whether the measurement is properly performed in the calibration, there may be a waste of time, such as a failure, e.g., poor contact at the time of the calibration, recognized upon a measurement of a test object after having completed the time-consuming calibration operation.
Patent document 1, Japanese Unexamined Patent Application Publication No. 6-34686, discloses a method for calibrating a network analyzer having two test terminals to be connected to a test object via a strip line. That is, a first calibration measurement is made to measure transmission and reflection parameters of the microstrip line whose propagation constant is unknown, which is connected between the two test terminals in a reflection-free manner. Three further calibration measurements are made using the same line and three calibration standards realized with reflection-symmetric and reciprocal discontinuous objects disposed at three different positions on the line.
That is, the three types of standards are realized by changing the state of the transmission line to three states. This way, the standards are connected only once. With this method, compared with TRL calibration, the number of times the standards are connected is reduced, and hence measurement errors in the calibration operation are reduced in number.
However, in the actual measurement of a test object, it is necessary to remove the strip line employed as the standard, and to again connect a strip line (fixture) to which the test object can be connected. Needless to say, characteristics of a reconnected portion change, resulting in measurement errors.
It is practically difficult to connect the strip line between the two test terminals in a reflection-free manner. Reflection parameters of portions in which the test terminals are connected to the strip line may cause errors.
Measured values obtained by connecting a test object are not characteristics of the test object alone, but are composite characteristics of the test object and the strip line to which the test object is connected. It is thus impossible to measure the characteristics of the test object alone.